Do you mean overclocking turbo +4 also adds 10-12% on intel iGPU?
If what you are saying is true couldn't we OC the baseclock on the i5 2550k (no igp)? Just curious
Well yes, and no, ish to both!
I'm going to use a music metaphor to hopefully explain this:
It's like an orchestra - all the components are have separate parts to play with different rhythms but they all interact at the same speed (the tempo directed by the conductor).
Everything is led by the cpu's baseclock - it is the speed at which the system talks to each other. (This is also known as the QPI)
If you change the baseclock then you change the speed at which all the components have to talk to each other. On intel i3, i5 and i7 processors, generally speaking, this could include the RAM speed, pcie speed, igpu speed and the cpu.
For whatever reason sandy doesn't like this. It prefers to alter the multiplier rather than the bclk. The relationship between the baseclock and multiplier is this:
The bclk is the speed at which the components report their work to one another - so 100mhz means that the components talk to each other 100 times a second.
The multiplier is the amount of work the component does within each reporting period. So if a component works at 10x multiplier it will do 10 bits of work within each 100mhz reporting period - so to the outside world looking in it is working at 100x10 = 1ghz.
I haven't owned a SB chip but certainly on Clarkdale you could disable the igpu or unlink it from the bclk along with the reducing the mulipliers on the cpu and ram to see how far you could push the bclk. I haven't seen a 2550k which OC'd on the bclk so I guess the rest of the links don't like it either.